Large language models (LLMs) predominantly employ decoder-only transformer architectures, necessitating the retention of keys/values information for historical tokens to provide contextual information and avoid redundant computation. However, the substantial size and parameter volume of these LLMs require massive GPU memory. This memory demand increases with the length of the input text, leading to an urgent need for more efficient methods of information storage and processing. This study introduces the Anchor-based LLM (AnLLM), which utilizes an innovative anchor-based self-attention network (AnSAN) and also an anchor-based inference strategy. This approach enables LLMs to compress sequence information into an anchor token, reducing the keys/values cache and enhancing inference efficiency. Experiments show that the AnLLM maintains comparable accuracy with up to 99% keys/values cache reduction and up to 3.5 times faster inference. Despite a minor compromise in accuracy, the AnLLM significantly improves computational efficiency and resource utilization, demonstrating the potential of the anchor-based attention approach in the context of LLMs for real-time inference in practical applications.
Session-based recommendation has gained increasing attention in recent years, with its aim to offer tailored suggestions based on users' historical behaviors within sessions. To advance this field, a variety of methods have been developed, with ID-based approaches typically demonstrating promising performance. However, these methods often face challenges with long-tail items and overlook other rich forms of information, notably valuable textual semantic information. To integrate text information, various methods have been introduced, mostly following a naive fusion framework. Surprisingly, we observe that fusing these two modalities does not consistently outperform the best single modality by following the naive fusion framework. Further investigation reveals an potential imbalance issue in naive fusion, where the ID dominates and text modality is undertrained. This suggests that the unexpected observation may stem from naive fusion's failure to effectively balance the two modalities, often over-relying on the stronger ID modality. This insight suggests that naive fusion might not be as effective in combining ID and text as previously expected. To address this, we propose a novel alternative training strategy AlterRec. It separates the training of ID and text, thereby avoiding the imbalance issue seen in naive fusion. Additionally, AlterRec designs a novel strategy to facilitate the interaction between the two modalities, enabling them to mutually learn from each other and integrate the text more effectively. Comprehensive experiments demonstrate the effectiveness of AlterRec in session-based recommendation. The implementation is available at https://github.com/Juanhui28/AlterRec.
Person re-identification (re-ID) continues to pose a significant challenge, particularly in scenarios involving occlusions. Prior approaches aimed at tackling occlusions have predominantly focused on aligning physical body features through the utilization of external semantic cues. However, these methods tend to be intricate and susceptible to noise. To address the aforementioned challenges, we present an innovative end-to-end solution known as the Dynamic Patch-aware Enrichment Transformer (DPEFormer). This model effectively distinguishes human body information from occlusions automatically and dynamically, eliminating the need for external detectors or precise image alignment. Specifically, we introduce a dynamic patch token selection module (DPSM). DPSM utilizes a label-guided proxy token as an intermediary to identify informative occlusion-free tokens. These tokens are then selected for deriving subsequent local part features. To facilitate the seamless integration of global classification features with the finely detailed local features selected by DPSM, we introduce a novel feature blending module (FBM). FBM enhances feature representation through the complementary nature of information and the exploitation of part diversity. Furthermore, to ensure that DPSM and the entire DPEFormer can effectively learn with only identity labels, we also propose a Realistic Occlusion Augmentation (ROA) strategy. This strategy leverages the recent advances in the Segment Anything Model (SAM). As a result, it generates occlusion images that closely resemble real-world occlusions, greatly enhancing the subsequent contrastive learning process. Experiments on occluded and holistic re-ID benchmarks signify a substantial advancement of DPEFormer over existing state-of-the-art approaches. The code will be made publicly available.
We propose BERT4FCA, a novel method for link prediction in bipartite networks, using formal concept analysis (FCA) and BERT. Link prediction in bipartite networks is an important task that can solve various practical problems like friend recommendation in social networks and co-authorship prediction in author-paper networks. Recent research has found that in bipartite networks, maximal bi-cliques provide important information for link prediction, and they can be extracted by FCA. Some FCA-based bipartite link prediction methods have achieved good performance. However, we figured out that their performance could be further improved because these methods did not fully capture the rich information of the extracted maximal bi-cliques. To address this limitation, we propose an approach using BERT, which can learn more information from the maximal bi-cliques extracted by FCA and use them to make link prediction. We conduct experiments on three real-world bipartite networks and demonstrate that our method outperforms previous FCA-based methods, and some classic methods such as matrix-factorization and node2vec.
Invariance learning algorithms that conditionally filter out domain-specific random variables as distractors, do so based only on the data semantics, and not the target domain under evaluation. We show that a provably optimal and sample-efficient way of learning conditional invariances is by relaxing the invariance criterion to be non-commutatively directed towards the target domain. Under domain asymmetry, i.e., when the target domain contains semantically relevant information absent in the source, the risk of the encoder $\varphi^*$ that is optimal on average across domains is strictly lower-bounded by the risk of the target-specific optimal encoder $\Phi^*_\tau$. We prove that non-commutativity steers the optimization towards $\Phi^*_\tau$ instead of $\varphi^*$, bringing the $\mathcal{H}$-divergence between domains down to zero, leading to a stricter bound on the target risk. Both our theory and experiments demonstrate that non-commutative invariance (NCI) can leverage source domain samples to meet the sample complexity needs of learning $\Phi^*_\tau$, surpassing SOTA invariance learning algorithms for domain adaptation, at times by over $2\%$, approaching the performance of an oracle. Implementation is available at https://github.com/abhrac/nci.
Existing methods for scene text detection can be divided into two paradigms: segmentation-based and anchor-based. While Segmentation-based methods are well-suited for irregular shapes, they struggle with compact or overlapping layouts. Conversely, anchor-based approaches excel for complex layouts but suffer from irregular shapes. To strengthen their merits and overcome their respective demerits, we propose a Complementary Proposal Network (CPN) that seamlessly and parallelly integrates semantic and geometric information for superior performance. The CPN comprises two efficient networks for proposal generation: the Deformable Morphology Semantic Network, which generates semantic proposals employing an innovative deformable morphological operator, and the Balanced Region Proposal Network, which produces geometric proposals with pre-defined anchors. To further enhance the complementarity, we introduce an Interleaved Feature Attention module that enables semantic and geometric features to interact deeply before proposal generation. By leveraging both complementary proposals and features, CPN outperforms state-of-the-art approaches with significant margins under comparable computation cost. Specifically, our approach achieves improvements of 3.6%, 1.3% and 1.0% on challenging benchmarks ICDAR19-ArT, IC15, and MSRA-TD500, respectively. Code for our method will be released.
Wireless capsule endoscopy (WCE) is a non-invasive diagnostic procedure that enables visualization of the gastrointestinal (GI) tract. Deep learning-based methods have shown effectiveness in disease screening using WCE data, alleviating the burden on healthcare professionals. However, existing capsule endoscopy classification methods mostly rely on pre-defined categories, making it challenging to identify and classify out-of-distribution (OOD) data, such as undefined categories or anatomical landmarks. To address this issue, we propose the Endoscopy Out-of-Distribution (EndoOOD) framework, which aims to effectively handle the OOD detection challenge in WCE diagnosis. The proposed framework focuses on improving the robustness and reliability of WCE diagnostic capabilities by incorporating uncertainty-aware mixup training and long-tailed in-distribution (ID) data calibration techniques. Additionally, virtual-logit matching is employed to accurately distinguish between OOD and ID data while minimizing information loss. To assess the performance of our proposed solution, we conduct evaluations and comparisons with 12 state-of-the-art (SOTA) methods using two publicly available datasets. The results demonstrate the effectiveness of the proposed framework in enhancing diagnostic accuracy and supporting clinical decision-making.
Geometry problem solving has always been a long-standing challenge in the fields of automated reasoning and artificial intelligence. This is the fifth article in a series of our works, we built a neural-symbolic system to automatically perform human-like geometric deductive reasoning. The symbolic part is a formal system built on FormalGeo, which can automatically perform geomertic relational reasoning and algebraic calculations and organize the solving process into a solution hypertree with conditions as hypernodes and theorems as hyperedges. The neural part, called HyperGNet, is a hypergraph neural network based on the attention mechanism, including a encoder to effectively encode the structural and semantic information of the hypertree, and a solver to provide problem-solving guidance. The neural part predicts theorems according to the hypertree, and the symbolic part applies theorems and updates the hypertree, thus forming a Predict-Apply Cycle to ultimately achieve readable and traceable automatic solving of geometric problems. Experiments demonstrate the correctness and effectiveness of this neural-symbolic architecture. We achieved a step-wised accuracy of 87.65% and an overall accuracy of 85.53% on the formalgeo7k datasets. The code and data is available at https://github.com/BitSecret/HyperGNet.
Recently, Large Language Models (LLMs) have been widely studied by researchers for their roles in various downstream NLP tasks. As a fundamental task in the NLP field, Chinese Grammatical Error Correction (CGEC) aims to correct all potential grammatical errors in the input sentences. Previous studies have shown that LLMs' performance as correctors on CGEC remains unsatisfactory due to its challenging task focus. To promote the CGEC field to better adapt to the era of LLMs, we rethink the roles of LLMs in the CGEC task so that they can be better utilized and explored in CGEC. Considering the rich grammatical knowledge stored in LLMs and their powerful semantic understanding capabilities, we utilize LLMs as explainers to provide explanation information for the CGEC small models during error correction to enhance performance. We also use LLMs as evaluators to bring more reasonable CGEC evaluations, thus alleviating the troubles caused by the subjectivity of the CGEC task. In particular, our work is also an active exploration of how LLMs and small models better collaborate in downstream tasks. Extensive experiments and detailed analyses on widely used datasets verify the effectiveness of our thinking intuition and the proposed methods.
Previous studies in predicting crash risk primarily associated the number or likelihood of crashes on a road segment with traffic parameters or geometric characteristics of the segment, usually neglecting the impact of vehicles' continuous movement and interactions with nearby vehicles. Advancements in communication technologies have empowered driving information collected from surrounding vehicles, enabling the study of group-based crash risks. Based on high-resolution vehicle trajectory data, this research focused on vehicle groups as the subject of analysis and explored risk formation and propagation mechanisms considering features of vehicle groups and road segments. Several key factors contributing to crash risks were identified, including past high-risk vehicle-group states, complex vehicle behaviors, high percentage of large vehicles, frequent lane changes within a vehicle group, and specific road geometries. A multinomial logistic regression model was developed to analyze the spatial risk propagation patterns, which were classified based on the trend of high-risk occurrences within vehicle groups. The results indicated that extended periods of high-risk states, increase in vehicle-group size, and frequent lane changes are associated with adverse risk propagation patterns. Conversely, smoother traffic flow and high initial crash risk values are linked to risk dissipation. Furthermore, the study conducted sensitivity analysis on different types of classifiers, prediction time intervalsss and adaptive TTC thresholds. The highest AUC value for vehicle-group risk prediction surpassed 0.93. The findings provide valuable insights to researchers and practitioners in understanding and prediction of vehicle-group safety, ultimately improving active traffic safety management and operations of Connected and Autonomous Vehicles.